deepseek-v4.1IMAGE
DeepSeek V4.1 Flash on M5 Ultra via MLX: 46.3 tok/s on 1M-context long text
Running DeepSeek-V4.1-Flash-1017 locally on an M5 Ultra 256GB with an MLX recipe, hitting 46.3 tok/s on long 1M-context text and 50 tok/s on normal prose.
KimSunIl
@ksuniriArchived 1 day agoOriginally posted 2026.10.27
KimSunIl @ksuniriOriginally posted 2026.10.27
##### Development Environment ##### Spec : M5 Ultra 256g / 4TB Recipe : @BobClawblaw , https://github.com/BobClawblaw/deepseek-v41-flash-mlx-m5-ultra Model : DeepSeek-V4.1-Flash-1017 https://huggingface.co/BobClawblaw/DeepSeek-V4.1-Flash-1017-TensorFold-e2e3-p4-h8 CTX : 1M Decode 1M 장문 -> 46.3 tok/s , Normal Prose -> 50 tok/s 결론 : M5 Ultra 512g 도 소유하고 싶다^^
x.com/ksuniri/status/2114905982321254683Archived 2026.10.27 12:07